HMNE SFT Adapter

• Base model: Qwen/Qwen2.5-3B-Instruct
• Dataset repo: Rehangmu1710/hmne-sft-v1
• Train split: train (2700 rows)
• Validation split: validation (300 rows)
• Dataset format: text
• Max sequence length: 1536
• Epochs: 1.0
• Learning rate: 0.0002
• Batch size: 4
• Gradient accumulation: 4